Suzanne Nossel is chief executive officer of PEN America and the author of “Dare to Speak: Defending Free Speech for All.”
This week Tesla CEO and Silicon Valley titan Elon Musk struck a deal to buy Twitter for about $44 billion and take the company private, an acquisition that could transform the social-media landscape and the daily experiences of millions of users around the world.
“Free speech is the bedrock of a functioning democracy, and Twitter is the digital town square where matters vital to the future of humanity are debated,” Musk said in a statement announcing the acquisition. He made clear that he aims to improve the product by vigorously promoting freedom of speech on the platform.
Musk is right to call for more aggressive and effective defenses of free speech online. But to the extent his vision of a freewheeling public forum implies the wholesale repeal of the rules that currently govern Twitter and other social media platforms, Musk will learn the hard way that there is no return to a mythic online Eden where all forms of speech flourish in miraculous harmony.
On the contrary, for Twitter to survive as a viable forum to share information, surface ideas and spark controversy, robust guardrails in the form of intensive content moderation are essential. Without them the platform would be overrun with spam, harassment, doxing, vitriol, obscenity, hucksterism, conspiracy theories, quackery, propaganda and calls to real world violence. Tearing up the rulebook would bury Twitter users under an avalanche of digital detritus, driving eyeballs and worthwhile conversations elsewhere.
That’s not to say that Musk can do nothing to put his thumb on the side of augmented safeguards for free expression on a retooled Twitter. There is certainly room to review Twitter’s usage rules and investigate areas where the company may be overly aggressive in moderation practices. Musk’s reported concern that Twitter’s content controls may reflect the biases of its San Francisco-based staff and lean too heavily toward policing conservative speech should be researched to see if hard data substantiates a problem; research studies to date have suggested that the purported problem does not exist.
Beyond such steps, if Musk wants to truly reshape the landscape for free speech on social media, he would do best to recognize that robust content moderation is here to stay. Musk should home in on addressing an unavoidable outgrowth of companies’ active monitoring of platforms for harmful speech: namely a persistent stream of so-called “false positives” — in other words, posts that trigger review and takedown but don’t actually violate platform rules.
Such false positives may involve satirical references to speech that would otherwise be bullying, the empowering reclamation of slur terms, or simply posts that are taken out of context and misinterpreted by either human or automated moderation schemes. The muzzling of legitimate discourse online is often the product of imperfect systems processing reams of posts in real time, and erring on the side of removal for posts that, with more careful review, would be judged not to defy platform rules.
The problem of false positive takedowns is significant. Due both to the urgency of addressing COVID-related misinformation and the practical constraints of lockdown, the pandemic accelerated social media companies’ reliance on automated content moderation. But automated tools are at best a blunt instrument. Twitter has drawn scrutiny for restricting the accounts of writers who’ve reported on the expulsion of Palestinians from their homes in East Jerusalem, human-rights activists in Pakistan, and users swept up in the enforcement of a new policy after a wave of “coordinated and malicious reports” by far-right activists and white supremacists targeting journalists and researchers who track extremism. Data from other platforms offer a glimpse into the extent of the false-positives problem: TikTok figures suggest that about 5 percent of videos removed for rules violations are eventually restored, amounting to more than 4 million videos per quarter.
Inevitably, because of the scale of the content to sift and the limits of artificial intelligence, more aggressive automated moderation methods will be overly punitive – flagging, demoting or deleting some content without good reason. The problem of false positives will only intensify with the adoption of new content moderation regulations in Europe that impose stiff penalties for platforms that fail to swiftly remove content that violates local laws.
Compounding the problem is a lack of transparency about the companies’ methods. Platforms generally offer only terse notices when content vanishes, often failing to specify precisely what rule was broken or how. When users appeal such a determination, resolution can take weeks, by which time whatever news or information value a post might have will almost certainly have evaporated. Facebook’s Oversight Board reviews a relatively small number of borderline policy calls on disputed content each year, but the process takes many months, and offers no recourse for most ordinary users who have no clue why their posts disappeared or accounts were disabled, or what can be done about it.
If Musk wants to tip the scales in favor of free speech, he could start by augmenting Twitter with a robust corrective to the problem of false positives. Specifically, he could work with the company’s engineers to put in place a content defense service that arms users with the ability to challenge dubitable enforcements in real-time. Such a service would work to quickly rectify instances in which legitimate posts that violate neither the law nor company content guidelines are nonetheless erroneously removed or suppressed. By establishing a powerful failsafe to guard against overly zealous content moderation, Twitter could maintain and refine its methods for addressing the genuine harms of online content while still achieving far greater protection for free speech.
With a reliable, universal, publicly accountable — and adequately resourced — system to ensure that erroneous content removal decisions can be quickly reviewed and – where appropriate — reversed, more aggressive automated moderation would pose less peril for speech. Such a scheme could borrow from related mechanisms that are well-established in other realms, including consumer payments, package tracking and the courts. Alert to fraud, credit card providers have become quick to disable charges that trigger a flag by deviating from past customer usage patterns or meeting other criteria that can indicate fraud. Often, such declines are unwarranted; the cardholder has simply traveled to a new city or splurged on a big purchase. When that happens, some hold time and explanation is normally all it takes to put the charge through.
The same should be true for erroneously flagged content. If a post is deleted or demoted, or an account disabled, the user should be immediately notified by phone, email or text and directed to contact a hotline if they believe the determination is wrongful. Users should be able to avail themselves, around the clock and in a wide array of languages, of the assistance of a trained and empowered staff that can examine straightforward cases and, where justified, be authorized to reverse unwarranted decisions “while you wait.”
So, if a contextualized look at a post reveals that it is satirical or otherwise non-mendacious, content defense staff should be empowered to restore it in real time. Where the matter cannot be resolved instantaneously, a tracking number should be assigned allowing the user to monitor exactly where their case stands as it moves through levels of review. Musk could invest enough in the service to minimize wait times and guarantee that those who believe their expressive rights have been breached can get help quickly. The creation of an escalation channel could allow for expedited appeals in cases where delaying restoration of content or accounts could be harmful, for example in the case of journalists covering conflict or protests, or documentation of human rights abuses.
By equipping users with expert assistance to contest dubious determinations, such a service would help balance Twitter’s drive to mitigate harm with both Musk’s – and the public’s – interest in keeping online expression as freewheeling as possible. By operating transparently, with periodic public reporting on categories of issues addressed and their resolution, the service could be a boon for Twitter users and the public at large, flouting predictions that the platform’s migration into private hands will only further impair accountability.
Musk is correct that our current systems of content moderation are deeply flawed. While a free-for-all vision for Twitter is not the answer, there are concrete steps the company could take to better uphold freedom of speech. A more robust, accessible, speedy and accountable approach– a more human approach– to user appeals to challenge content decisions would be one major step toward creating a public sphere in which all ideas can get the airing they deserve.
Suzanne Nossel currently serves as the Chief Executive Officer of PEN America, the leading human rights and free expression organization, and she is author of Dare to Speak: Defending Free Speech for All. Since joining in 2013, she has doubled the organization’s staff, budget, and membership, spearheaded the unification with PEN Center USA in Los Angeles and the establishment of a Washington, D.C. office, and overseen groundbreaking work on free expression in Hong Kong and China, Myanmar, Eurasia, and the United States. She is a leading voice on free expression issues in the United States and globally, writing and being interviewed frequently for national and international media outlets. Nossel is a magna cum laude graduate of both Harvard College and Harvard Law School. In 2021, Nossel was selected as a member of the Oversight Board, an independent body using human rights principles to adjudicate decisions on Facebook and Instagram.